the classified images, and 14 from the resampled images (at the aggregation levels
between 30 and 240 m) of the ETM+ and ASTER images. In doing so, the efficiency
of various spatial techniques in measuring and characterizing land covers in response
to scale and resolution change could be more thoroughly examined.
12.3 LANDSCAPE CHARACTERIZATION AT MULTIPLE SCALES
BY FRACTAL MEASUREMENT
Among the many FD algorithms, the most popular one seems to be the TP algorithm
due to its robustness (Emerson et al., 2005; Liang and Weng, 2013; Soille and
Rivest, 1996). Developed by Clarke (1986), this method employs the idea of
interpreting images from a three-dimensional perspective by viewing all the pixel
values as “elevation.” Taking the four adjacent pixels that make up the four corners
of a square and their mean value to build a vertical line which locates centrally, a
prism with four triangular facets is thus constructed. The method is implemented by
using these triangular facets as the basic tool to fill an image surface until the total
surface area is calculated. The same procedure is repeated for each step size that
equals the number of pixels on a side of the square. The regression model for this
FIGURE 12.2 Red bands of Landsat MSS75 (a), Landsat TM85 (b), Landsat TM95 (c),
Landsat ETM + 00 (d), AST01 (e), IKN01 subscene for residential area ( f), and IKN03
subscene for downtown area (g).
236
MULTISCALE FRACTAL CHARACTERISTICS OF URBAN
between 30 and 240 m) of the ETM+ and ASTER images. In doing so, the efficiency
of various spatial techniques in measuring and characterizing land covers in response
to scale and resolution change could be more thoroughly examined.
12.3 LANDSCAPE CHARACTERIZATION AT MULTIPLE SCALES
BY FRACTAL MEASUREMENT
Among the many FD algorithms, the most popular one seems to be the TP algorithm
due to its robustness (Emerson et al., 2005; Liang and Weng, 2013; Soille and
Rivest, 1996). Developed by Clarke (1986), this method employs the idea of
interpreting images from a three-dimensional perspective by viewing all the pixel
values as “elevation.” Taking the four adjacent pixels that make up the four corners
of a square and their mean value to build a vertical line which locates centrally, a
prism with four triangular facets is thus constructed. The method is implemented by
using these triangular facets as the basic tool to fill an image surface until the total
surface area is calculated. The same procedure is repeated for each step size that
equals the number of pixels on a side of the square. The regression model for this
FIGURE 12.2 Red bands of Landsat MSS75 (a), Landsat TM85 (b), Landsat TM95 (c),
Landsat ETM + 00 (d), AST01 (e), IKN01 subscene for residential area ( f), and IKN03
subscene for downtown area (g).
236
MULTISCALE FRACTAL CHARACTERISTICS OF URBAN